Special colloquium on individual differences in language attainment and their causes
Wednesday 14 June 2017
Sponsored by Language Learning
Organizers: Ewa Dabrowska and Sible Andringa
Most language acquisition researchers assume, either implicitly or explicitly, that all first language learners converge on the same grammar. This outcome contrasts sharply with the outcome of L2 acquisition, which is characterized by large individual differences, particularly in adult learners. Furthermore, adult learners rarely, if ever, attain native-like competence. These differences between first and second language acquisition are often attributed to a biologically determined critical period.
However, there is growing evidence that native speaker convergence is a myth: there are, in fact, considerable individual differences in adult L1 speakers’ linguistic competence (Dąbrowska 2012, Farmer et al. 2012, Hulstijn 2015, Misyak and Christiansen 2012) . These are attributable partly to differences in experience, in particular education and print exposure, and partly to learner internal factors such as statistical learning abilities, IQ, metalinguistic abilities, and need for cognition. This has important implications for second language research. First, the vast majority of ultimate attainment studies use highly educated participants (Andringa 2014). A very different picture emerges from studies which use a native control group which includes lower SES speakers, which often show many more L2 learners performing within the native range (Dąbrowska and Street 2006, Hulstijn 2015). It is unclear whether this is due to the fact that age effects are at least partly attributable to education in the L2, or whether it means that critical period effects also apply in first language acquisition. Secondly, the factors that predict differences in L1 attainment are well known to be involved in second language learning. Explicitly comparing their effects and interactions in both populations thus opens a whole new field of research.
This colloquium examines a variety of factors that determine the outcome of language acquisition, including both environmental factors (amount and type of input and the context in which learning takes place) and learner internal factors (IQ, language aptitude, implicit learning abilities, affective factors), using data from different populations of learners (monolingual v. bilingual, children v. adults, younger v. older learners, highly educated v. less educated v. illiterate learners). Some contributions will investigate the effects on global measures of proficiency, while others will examine differential effects on specific aspects of language, and even specific constructions, thus providing insights into the cognitive processes underlying language by mapping patterns of strengths and weaknesses in particular areas of language to individual differences in experience, motivation, and abilities.
10.30 Jan Hulstijn: A usage-based, probabilistic view on the mental grammars of adult native speakers and end-state nonnative speakers: Conceptual and methodological challenges for measuring language proficiency
11.00 Ewa Dąbrowska: Experience, aptitude and individual differences in ultimate attainment: A comparison of L1 and L2 speakers
11.30 Catherine J. Doughty: Cognitive language aptitude: L2 ceiling and partial solution
12.00 Maja Curcic, Sible Andringa and Folkert Kuiken: Individual differences in L2 learners’ ability to process determiners online
15.00 David Singleton and Simone Pfenninger: Starting age overshadowed: The primacy of differential environmental and input effects on L2 attainment in an instructional context
15.30 Patricia J. Brooks and Vera Kempe: Individual Differences in language learning over the lifespan are inconsistent with the ‘Less is More’ Hypothesis
16.30 Wander Lowie and Marjolijn Verspoor: A dynamic and longitudinal perspective on Individual Differences
17.30 Martha Bigelow and Kendall King: Explaining different outcomes in language and literacy development among low-literate learners
Jan Hulstijn (J.H.Hulstijn@uva.nl)
University of Amsterdam
Comparing language proficiency in native and nonnative speakers (NSs and NNSs) is fraught with conceptual and methodological difficulties. In this presentation I propose to deconstruct the NS notion by characterizing NSs extralinguistically (i) along a biographical-environmental continuum from purely monolingualism to complete bi- or multilingualism and (ii) in terms of social-psychological attributes (age, level of education, profession, etc.). This would enable the study of linguistic cognition (language proficiency) as a function of variables along these two dimensions. I then propose a similar deconstruction of the NNS notion. I then raise the question of whether it would be possible to empirically establish the contents of lexical and grammatical cognition shared by all NSs (Miller & Weinert, 1998; Hulstijn, 2015). On the basis of illustrative examples from a small corpus of extemporaneous speech, produced by 98 adult NSs of Dutch, differing in age (18-76) and level of education and profession (High v. Low), I argue that the notion of shared linguistic grammatical cognition is neither empty nor trivially small. I then consider the question of how one could empirically establish, using techniques of probabilistic linguistics, whether a given end-state NNS has acquired NSs’ shared linguistic cognition. This approach, yielding a novel notion of nearnativeness/nativelikeness, might provide us with a new view on two classical issues: (i) To what extent is there an age-of-onset constraint in the attainment of nativelike L2 proficiency? and (ii) Is it possible for bilinguals to attain maximal command of both their languages?
Experience, aptitude and individual differences in ultimate attainment:
A comparison of L1 and L2 speakers
Ewa Dąbrowska (firstname.lastname@example.org)
Language acquisition researchers often assert that L1 learners converge relatively rapidly on the same grammar, while L2 learners do not. However, a number of recent studies have cast doubt on this assertion: there are, in fact, considerable individual differences in L1 ultimate attainment (see Dąbrowska 2015 for a recent review). Some, though not all, of these differences are related to education: while highly educated participants typically perform at ceiling, there is much more variability among less educated speakers.
This contribution compares the performance native speakers (N=60) and adult L2 learners (N=69) on tasks tapping proficiency in three linguistic domains: grammar, vocabulary and collocations. In addition, data was collected on four possible predictors of individual differences in linguistic attainment: education, nonverbal IQ, language aptitude, and print exposure. Both groups included participants from a variety social and educational backgrounds, with the sample approximating, in so far as possible, the demographics of the UK population.
As anticipated, the native group outperformed L2 speakers on all three language measures. However, the effect sizes varied according to area of linguistic knowledge: relatively small for grammar, moderately large for vocabulary, and very large for collocations. Moreover, there was considerable overlap between the scores in the two groups, particularly for grammar, where 74% of the nonnative participants fell within the native speaker range, and 49% performed above the native mean. These results are difficult to reconcile with critical period explanations of age effects. Regression analyses revealed some interesting differences between the groups in which nonlinguistic measures best predict performance on the language tasks. These findings suggest that L1 and L2 learners rely on the same cognitive mechanisms when learning language, albeit to different extents, and that the differences between the two populations are explainable in terms of the amount and type of linguistic experience, and possibly motivation.
Cognitive language aptitude: L2 ceiling and partial solution
Catherine J. Doughty (email@example.com)
University of Maryland
I begin with a brief discussion of the role of language aptitude in second language acquisition (SLA) within the context of other important individual differences, such as motivation and personality. Next, I will make and defend a claim that language aptitude is a stable trait which imposes a ceiling on post-critical SLA attainment and explains why some, but not all, adult learners attain very high second language (L2) ability. Then, I turn to specific language aptitude constructs and their operationalization in language aptitude batteries, considering some key measurement issues (e.g., reliability, influence of known languages, and fidelity to SLA processes). Given this background information, I present a model of multi-componential, cognitive language aptitude and review recent validity research on the High-Level Language Aptitude Battery (Hi-LAB) which instantiates the model. In the second part, I discuss aptitude-by- treatment interactions (ATI) in instructed SLA (ISLA) empirical research that matches pedagogic procedures to the strongest (in contrast to the weakest) cognitive language aptitude components for each individual learner, which optimizes L2 learning outcomes underneath the ceiling. I will briefly review ATI in ISLA research to date, mentioning the challenges of conducting such research, useful research designs, and the issue of ecological validity. Then will discuss in greater detail a three-year evaluation of a tailored language training program, including a description of the layers of tailoring (curricular and aptitude-based); development and use of aptitude profiles; and outcomes from multiple assessments, including global proficiency, fine-grained measures (linguistic correlates of proficiency), and job-relevant, task-based tests. Finally, I describe recent efforts to scale up aptitude-based, individualized instruction through classroom observations and materials review, leading to a guidebook for learners/instructors and by incorporating aptitude components in online language learning.
Individual differences in L2 learners’ ability to process determiners online
Maja Curcic (M.Curcic@uva.nl), Sible Andringa and Folkert Kuiken
University of Amsterdam
Several studies have demonstrated that L2 learners can use gender-marked determiners to predict upcoming nouns in nativelike ways, and that this ability depends on learners’ proficiency and L1 characteristics (e.g. Lew-Williams & Fernald, 2010; Dussias et al., 2013). This study focuses on individual differences in L2 determiner processing and aims to investigate the role of learner internal factors such as awareness and aptitude that might be responsible for individual differences.
Fifty adult Dutch native speakers received brief, auditory exposure to a language based on Fijian, which included gender-marked determiners. They were asked to learn the language by looking at images and listening to phrases and sentences describing the images. There was no explicit focus on grammar. Determiner processing was measured using the visual world paradigm with eye-tracking. Learners also did several aptitude tests measuring their non-verbal IQ, grammar inference ability, and sound recognition ability.
The results showed that learners as a group could process determiners online, but substantial individual differences were found. Awareness of gender distinction in the language was found to be a prerequisite for online processing. Only learners who reported this awareness showed online processing. Learners’ gender awareness was predicted by their performance on the grammar inference task. There were different degrees of awareness. Some learners developed strategic awareness and gained a speed advantage.
This study shows that learners differ substantially in their ability to process determiners online. The results imply that the failure to become aware may lead to the failure of learning. More broadly speaking, this might imply for both native and non-native learners, that learning sometimes no longer proceeds due to the fact that learners no longer develop awareness of particular structures, which often seems to be crucial for making progress.
Starting age overshadowed: The primacy of differential environmental and input effects on L2 attainment in an instructional context
David Singleton (DSNGLTON@tcd.ie) and Simone Pfenninger (firstname.lastname@example.org)
Trinity College and University of Salzburg
Despite the consistent, contrary findings of some 40 years of research, many educationalists still proclaim that there are clear linguistic advantages associated with beginning L2 instruction early. This assertion is again called seriously into question by the results of a five-year longitudinal study in Switzerland, in which we tested a range of oral and written English skills of over 500 secondary school students, who had all learned Standard German and French at primary school, but only half of whom had had English from third grade (age 8 onwards, the remainder having started five years later at secondary school. The findings suggest that age effects are overshadowed by other effects, yielding diverse outcomes according to individual differences (e.g. in relation to literacy skills), and according to contextual effects that mediate L2 outcomes (e.g. language environment at home, classroom effects and teaching approach).
A particular focus of this paper is on the interaction between person and context in quantitative and qualitative age research. The precise significance of starting age and biological age is difficult to determine, because we cannot disentangle effects of maturation from effects of age-related circumstances. The age factor involves numerous interrelated variables, having to do with learner orientation and experience, macro- and micro-contextual influences, and hierarchical nesting structures within a given sample.
From a theoretical and research perspective, these arguments cry out for empirical studies which take a more nuanced look at what underlies age effects in SLA and investigate how learning contexts shape processes of L2 development in individuals. In our paper we propose research methods that enable us to integrate individual-level and contextual-level data in order to assess the impact of context-varying factors in relation to individual differences variables – such as statistical models that can maximize the generalizability of the findings without sacrificing too much variability.
Individual Differences in Language Learning over the Lifespan are Inconsistent with the ‘Less is More’ Hypothesis
Patricia J. Brooks (Patricia.Brooks@csi.cuny.edu) and Vera Kempe (email@example.com)
College of Staten Island and University of Abertay
The ‘Less-Is-More’ hypothesis views the limited processing capacity of infants and young children as beneficial for language learning. One version of this hypothesis implicates limitations in working memory capacity and executive control in children’s propensity to chunk and regularize input. Another version suggests that children’s limited knowledge favors exploration of a wider hypothesis space. Thus, developmental increases in capacity or knowledge base are invoked to explain age-related decreases in language learning ability (i.e., closure of a critical period). We point out that associations between individual differences in processing capacity associated with working memory, executive control, non-verbal intelligence and motor abilities on the one hand and language learning aptitude on the other hand run in the opposite direction: throughout the lifespan ‘More is More.’ In support, we summarize the vast literature indicating that working memory and statistical learning abilities predict individual differences in language acquisition in children, language processing in adults, L2 acquisition in adults, and are associated with childhood language impairments. These effects are cumulative and substantive, and are consistent with the view that the end-state of language development is not uniform. As an example, we show that individual differences in sustained attention in infancy predict vocabulary size at 3 years, after controlling for many other factors. We also point to recent work using iterated language learning and computational modeling which suggests that some apparent advantages of childhood learners appear to be due to different initial biases, which are experience-based and not the result of capacity limitations. This leads to a more general developmental question that can be posed within the Bayesian framework about where initial priors come from and how their continuous updating may affect language development throughout the lifespan. We explore possible answers.
A dynamic and longitudinal perspective on Individual Differences
Wander Lowie and Marjolijn Verspoor
University of Groningen
The ever increasing body of research into individual differences in second language learning tends to focus on the extent to which success in second language learning can be predicted by a set of relevant learner characteristics like motivation, aptitude, and personality. Many studies have concentrated on a limited number of learner characteristics and have indicated the relative importance of some of these factors. Since it is generally observed that there are major differences among the achievement levels of second language learners, it is tempting to conduct massive and comprehensive group studies to identify the keys to success in second language learning. However, such studies may underestimate the dynamic nature of second language development in which all factors dynamically interact in the time domain, leading to language development as a highly individual process. This contribution will give an account of a Complex Dynamic Systems Theory (CDST) approach to individual differences and will elaborate on their dynamic interactions. It will be argued that group studies that determine the relation between individual differences and language achievement at one point in time can only tell half the story. To discover differences between individual learning trajectories, longitudinal studies are bound to provide us with valuable additional information about the process of development. We will report on a study of the longitudinal development of English L2 writing by Dutch learners in a high exposure context. We followed the 20 early learners by collecting weekly data for a period of 22 weeks. The development of general proficiency (holistically scored texts) and the development of the complexity in their writing samples was analyzed in the context of several potential individual differences, starting proficiency in L1 and L2, scholastic aptitude, out of school exposure, and motivation. In spite of the strong similarity between these learners in most of the factors studied and the similar progress in proficiency when holistically scored, noticeable differences were observed in their individual trajectories .
Explaining different outcomes in language and literacy development among low-literate learners
Martha Bigelow (firstname.lastname@example.org) and Kendall King
University of Minnesota
This presentation examines late-onset literacy among English learning adolescents, an under-explored area of multilingual literacy research. Educators’ accounts frequently report that becoming literate in a new language, for the first time, takes longer and might never produce outcomes similar to individuals who become literate as children. While there are multiple causes for why adolescents do not acquire print literacy as children, by far the most common reason is lack of sustained access to formal schooling. Regardless of the cause of interrupted or limited formal schooling (e.g., political unrest), it is common to encounter adolescents with little or no print literacy, and UN data suggest their numbers will increase given the rise in refugees and lengthy delays in resettlement. This population, however, is far from monolithic. There are many individual differences among learners and wide variability in their progress (Bigelow & King, 2016).
This study analyzes the cases of four Somali newcomers to the US, all with refugee backgrounds and histories of limited access to formal schooling. Participants ranged from no print literacy to very low levels of literacy. Data consist of documentation of their English literacy shortly after arrival and throughout their first year in school. Methods include analysis of artifacts (e.g., worksheets, dictations), video and audio tapes of class participation in literacy events (English and Somali), and monthly individual literacy assessments.
Findings include a description of the participants’ progress and analyzes results through the lens of the critical/sensitive period hypothesis (Singleton, 1995; Snow & Hoefnagel-Höhle, 1975) as well as research programs in individual differences (Ellis, 1994). These findings contribute to a small but growing body of empirical data on low-literate learners and suggest ways in which inclusion of these often-over-looked learners in applied linguistics benefits and deepens understanding of second language processes (Bigelow & Tarone, 2004).